Input data risks · Question 21 of 60
Which mitigations help against poisoned training data?
4 answers are correct. Medium
Show answer
Answer
- A Data provenance testing
- C Exploratory data analysis
- D A/B testing
- F Attacks as part of red teaming
Explanation
A/B testing, data provenance testing, exploratory data analysis and attacks as part of red teaming help against poisoned training data. Label correctness testing is the mitigation for poor labeling guidelines, ambiguous data and inaccurate annotation.